The title is bait, and I took it. Nate B Jones — whose AI News & Strategy Daily channel is one of the more consistently substantive things in my feed — put out a video called “Your Roadmap Is Why You’re Losing to AI-Native Teams”, and I went in braced for another Agile is dead eulogy. That’s not what this is. It’s something more useful, and in a couple of places it’s making an argument I want to have out loud.
His setup will sound familiar if you’ve read this blog lately: AI made it cheap to build almost anything, and most companies still ship at the old pace. The reason, he argues, isn’t that their models are worse than Anthropic’s or OpenAI’s. It’s what they have — or haven’t — moved out of meetings and documents and into working code.
The argument in one move
Jones’s core claim is that every organization runs on implicit rules. They live in meeting cadences, approval chains, templates, and the tone of voice someone uses when they say that’s not how we do it here. Those rules weren’t arbitrary — they were sensible responses to scarcity. Engineering time was expensive. Context was slow to share. Mistakes were costly. The conditions argued for the rules.
Then AI dissolved the scarcity, and the rules kept enforcing themselves out of pure inertia. His fix is blunt: an operator’s job now includes rewriting the implicit rules as explicit, machine-readable ones — and then moving the repeatable coordination they govern into code. Not a values poster. Constraints that run: calendar checks, Slack challenges, audit tools, hard blocks. He wrote fifteen “commandments” for his own operation and walks through them in clusters. An engineering-and-product cluster: don’t slow engineering down; product makes no roadmaps and doesn’t direct engineering time — instead it’s in the terminal daily, sitting with engineering and jamming. A design-and-documentation cluster: no meetings over an hour; documentation as code; design moves into the code, the SDK, and the UI. And a human cluster: stay flexible enough to deliver value like water over stone; build with a team; help along the folks who aren’t going as fast, and learn from the ones who are.
The line that stuck with me is from the companion essay: a rule that lives only in the head of your longest-tenured director does not exist, as far as your stack is concerned. That’s the legibility problem in one sentence. Your agents can’t follow a rule nobody wrote down, and increasingly, neither can your people.
Where it lands
This rhymes hard with things I’ve been circling here for months, which is probably why I liked it.
I argued in The Patterns Didn’t Go Away that the quality bar didn’t move — it just lost its free ride, and now it has to be stated out loud on every change. Jones takes that a step further than I did: don’t just say the standard out loud, encode it where the work happens, so it gets enforced without a senior person spending judgment on it every time. And when I wrote about turning backlog items into contracts for AI agents, I was making his argument at the scale of a single ticket. He’s making it at the scale of the whole operating system.
The sharpest tool in the video is what he calls the enforcement ladder: a rule can be a value, an instruction, a reminder, a hard block, or a decision explicitly owned by a human — and you choose the rung deliberately. Most organizations I’ve seen only have the two ends: aspirational values nobody checks, and escalations to a person. The three middle rungs are where the leverage is, and they’re exactly the rungs that AI made cheap to build. His example of an audit tool that enforces the kill monthly meetings commandment is the whole thesis in miniature — coordination that used to cost a recurring hour of ten people’s attention, replaced by a check that runs.
Even the bait holds up. “Kill the roadmap” turns out to mean roughly what I meant by opportunity backlogs over feature backlogs: when building is cheap, a committed list of outputs is a liability, and the roadmap’s real functions — alignment, sequencing, saying no — get absorbed by the other rules. He’s explicit that the prohibition only works because the rest of the system picks up the load. Which brings me to the part where I got off the bus.
Where I push back
Jones insists the fifteen commandments work as one system, and that partial adoption is where this fails — you can’t cherry-pick the rule that feels easiest. His example is the org that stops holding roadmap meetings but never gets product into the terminal: the coordination is gone, the replacement never got installed, and what’s left is chaos wearing an AI-native costume. As a description of the design, he’s right, and he’s honest about it in a way most playbook-sellers aren’t. He even warns that copying somebody else’s constitution is how you end up with a document that enforces nothing.
But “every rule has to move together” is a big-bang change program, and I just spent a whole post arguing that organizations have a speed limit on absorbing exactly this kind of change. Fifteen simultaneous rewrites of how people meet, document, design, and decide is a lot of Everett Rogers’ adoption curve to swallow in one gulp — and pushing past that limit doesn’t accelerate adoption, it manufactures resistance. The failure mode he’s worried about (cherry-picking) and the failure mode I’m worried about (rejection of the transplant) are both real. The way through isn’t all-at-once; it’s sequencing the rules so each one, once absorbed, makes the next one cheaper to adopt — and being honest that the interlocking design means you haven’t captured the full value until the set is in. That’s a harder message than move it all together, but it’s the one that survives contact with an actual org chart.
The takeaway worth stealing
Don’t steal the commandments. Steal the method. For any rule your org enforces by habit: name the behavior, find the scarcity that originally justified it, ask whether that scarcity still exists, and if it doesn’t, rewrite the rule as a testable constraint and pick its rung on the ladder. That exercise costs you an afternoon, and it’s the part of the video I expect to still be using a year from now.
The photography analogy Jones opens with is the right frame to leave on: when taking a picture became free, photography didn’t die — judgment about which picture to take became the entire job. Building is going the same way. His bet is that the judgment belongs in code, where it compounds. Mine is that it gets there at the speed your people can absorb it, and not one commandment faster. Both things can be true. Watch the video and argue with us.
— Roger
Further reading
- Jones, Nate B. “Your Roadmap Is Why You’re Losing to AI-Native Teams.” AI News & Strategy Daily, YouTube — the video under review.
- Jones, Nate B. “AI-Native Companies Run on Code.” Nate’s Newsletter, Substack — the companion guide, with the full ruleset, the enforcement ladder, and the five-question worksheet.
Leave a comment